Remove the contributor locks

[ibaryshnikov]

Contributor's logic heavily relies on acquiring the locks. The lock itself signals that the contributor is working on a specific task at the moment. And the task is a combination of chunk id and contribution id. We have a lot of logic related to handling contributor locks. I've made an experiment moving the verifier lock logic to LRU cache on the application side. It allowed to remove the verifier lock logic from phase1-coordinator and improved overall stability of the service. We can expect the same outcome for the contributors: improved stability and better user experience.

Key parts of the change:

• remove the contributor locks in phase1-coordinator
• use LRU cache for assigned tasks in coordinator web service
• revisit contributor drop logic
• use assigned_tasks and completed_tasks in ParticipantInfo to choose what's the next task to work on. Remove pending_tasks, disposing_tasks and disposed_tasks